Hydrophobic nanoparticles improve permeability ofcell-encapsulating poly(ethylene glycol) hydrogelswhile maintaining patternabilityWonjae Leea, Nam-Joon Chob, Anming Xiongb, Jeffrey S. Glennb,c,1, and Curtis W. Frankd,2

aDepartment of Mechanical Engineering, Stanford University, Stanford, CA 94305; bDepartment of Medicine, Division of Gastroenterology andHepatology, School of Medicine, Stanford University, Stanford, CA 94305; cVeterans Administration Medical Center, Palo Alto, CA 94304; anddDepartment of Chemical Engineering, Stanford University, Stanford, CA 94305

Edited by Chad A. Mirkin, Northwestern University, Evanston, IL, and approved October 13, 2010 (received for review April 19, 2010)

Cell encapsulating poly(ethylene glycol) hydrogels represent apromising approach for constructing 3D cultures designed tomore closely approximate in vivo tissue environment. Improvedstrategies are needed, however, to optimally balance hydrogelpermeability to support metabolic activities of encapsulated cells,while maintaining patternability to restore key aspects of tissuearchitecture. Herein, we have developed one such strategy incorpo-rating hydrophobic nanoparticles to partially induce looser cross-linking density at the particle-hydrogel interface. Strikingly, ournetworkdesignsignificantly increasedhydrogel permeability,whileonly minimally affecting the matrix mechanical strength or prepo-lymer viscosity. This structural advantage improved viability andfunctions of encapsulated cells and permitted micron-scale struc-tures to control over spatial distribution of incorporated cells. Weexpect that this design strategy holds promise for the developmentof more advanced artificial tissues that can promote high levels ofcell metabolic activity and recapitulate key architectural features.

tissue engineering ∣ cell encapsulation ∣ hydrogel network properties ∣hepatitis C virus

Engineered tissues are expected to help overcome shortcom-ings associated with standard in vitro 2D culture and tradi-

tional tissue replacement therapies and provide an improvedplatform to study various biological assays (1). Several typesof cells cultured as two-dimensional monolayers are known todedifferentiate and lose their original functions due to disruptionof their normal microenvironments (2). Thus, it has been a majorchallenge in the field of tissue engineering to provide structuraland functional support equivalent to native tissues (3). Use ofhydrogels as a scaffold material for engineered tissues has beenof increasing interest due to their structural similarity to theextracellular components in the body (4). PEG is one of the mostextensively utilized hydrogels because its network structure canbe easily modified to mimic critical aspects of the original micro-environments (5). Various approaches to design PEG networksfor improved phenotype stability of encapsulated cells haveinvolved conjugating other biologically active factors to the poly-mer network and incorporating degradable linkages (6). Anotheradvantage of PEG is that use of photolithography or microfluidicprocessing allows fabrication of microarchitectures that poten-tially recapitulate key aspects of tissue architecture to guide cells’behavior with respect to morphology, cytoskeletal structure, andfunctionality (7).

Beyond the restoration of its original microarchitecture, athree-dimensional matrix of an engineered tissue needs to providesufficient permeability to support metabolic activities, provide forunimpeded transport of large macromolecules, and permit multi-ple cell-type interactions that are normally present in most origi-nal tissues (8). Even though the matrix diffusion condition couldbe gradually improved by promoting angiogenesis or conjugatingdegradable linkages (9), it is necessary to allow sufficient diffusionof substances indispensable to cell survival during the period

immediately after cells are encapsulated (6). In addition, becausemany scaffolds for tissue engineering initially need to fill a spaceand provide a framework for the replaced tissue, the mechanicalproperties of the material are important (6). However, it is a chal-lenge for hydrogel-based scaffolds to allow sufficient permeabilitywhile maintaining mechanical properties sufficient for generatingand sustaining reconstructed tissue architectures. In many poly-meric materials, improved permeability of networks has beenobtained by increasing the distance between consecutive cross-links (10). This comes, however, at a cost of decreasedmechanicalstrength in the resulting hydrogel network. Nevertheless, the tra-ditional approach also results in increased viscosity of the liquidphase prepolymer, which retards flow of the prepolymer solutionthrough micro-scale mold structures and also leads to decreasedmechanical strength of the cured hydrogel. There have beenattempts to address the relationship between the traditionalpolymeric parameters and the behavior of encapsulated cells inPEG networks (11). We recently determined that cell encapsula-tion leads to the formation of network defects and these networkdefects are up to orders of magnitude larger than the possibledistance range between the network cross-links (12). In this work,we suggest that the level of these network defects is an appropriatespatial feature for tuning network permeability and patternability.We therefore attempted to augment network defects by incor-porating hydrophobic poly(lactic-co-glycolic acid) (PLGA) nano-particles (NPs) within the hydrophilic PEG network. This modi-fication improved hydrogel permeability with minimum loss ofpatternability. Our network design led to improved viability andfunction of encapsulated human liver-derived cells and has thestructural advantage of supporting micron-scale architecturesfor control of the spatial distribution of incorporated cells.

Results and DiscussionAugmentation of Network Defects by Incorporating HydrophobicNanoparticles Improves the Phenotype Stability of Encapsulated Cells.Because the viability of encapsulated liver-derived cells has beenshown to be susceptible to the encapsulating matrix diffusion con-

Author contributions: W.L., N.-J.C., J.S.G., and C.W.F. designed research; W.L. and A.X.performed research; W.L. and N.-J.C. analyzed data; and W.L., N.-J.C., J.S.G., and C.W.F.wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.1To whom correspondence may be addressed at: Division of Gastroenterology andHepatology, School of Medicine, Stanford University, Center for Clinical Science ResearchBuilding, Room 3110, 269 Campus Drive, Stanford, CA 94305. E-mail: jeffrey.glenn@stanford.edu.

2To whom correspondence may be addressed at: Department of Chemical Engineering,Stanford University, 381 North-South Mall, Stauffer III, Stanford, CA 94305. E-mail: curt.frank@stanford.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1005211107/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1005211107 PNAS Early Edition ∣ 1 of 6

ENGINEE

RING

APP

LIED

BIOLO

GICAL

SCIENCE

S

Dow

nloa

ded

by g

uest

on

July

20,

202

1

dition (13), we utilized human liver-derived cells as a modelto evaluate the structure of cell-encapsulating PEG networks.Immortalized Huh 7.5 cells were chosen to optimize cell-encap-sulating conditions prior to studies with human primary hepato-cytes. Based on this scheme, we expanded upon recent workon the scale of network defects within a cell-encapsulating PEGhydrogel (12). In this previous work, we first encapsulated cellsinto PEG matrices and added virus particles into the cell culturemedium (12). We then observed that encapsulated liver-derivedcells could be infected with hepatitis C and pseudotyped lenti-viruses, and the progeny of the viruses could be recovered fromthe media supernatants (12). The sizes of the hepatitis C virus(HCV) and the lentivirus particles are 50 nm and 100 nm, respec-tively. However, when the possible distances between consecutivecross-links are roughly estimated from uniformly cross-linkednetworks of 3.4 k or 8 k PEG, they cannot exceed 5 nm (14). Wesuggested that the cell-encapsulating PEG networks had numer-ous network defects large enough for the virus particles to suc-cessfully penetrate and infect the cells (12). Therefore, networkdefects appear to provide a major pathway for solute diffusion,such that the overall permeability would not be expected tobe substantially improved by designing the network to haveincreased distances between the consecutive cross-links. To testthis hypothesis, we measured cell viability in samples with eitherincreased molecular weight of the polymer (e.g., 8 k vs. 3.4 k)or decreased solid content (e.g., 20% vs. 10%) as compared tothe 20% 3.4 k PEG reference network. As shown in Fig. 1A,the results imply that all of the tested networks appeared tohave similar levels of network defects to that of the referenceas assessed by cell viability. The decreased viability in samplesprepared with a further increase in solid content from 20% 3.4 kPEG (Fig. 1B) seems that fewer network defects were producedduring polymerization, and the distances between consecutivecross-links determined the overall network permeability. Basedon this understanding of the cell-encapsulating PEG network,we sought to augment the level of network defects to furtherimprove the permeability.

It has been reported that hydrogels cured in contact withhydrophobic surfaces exhibit characteristics attributed to lowercross-linking density than hydrogels cured on hydrophilic glasses(15). The major driving force for inhomogeneous gelation adja-cent to a hydrophobic surface is thought to be the high interfacialenergy between the hydrophobic substrate and the aqueous poly-merizing solution (16). Thus, in an attempt to further disruptthe PEG network structure so as to increase its permeability,we incorporated hydrophobic NPs into the hydrogel matrix. Byanalogy, the large interfacial zone between the hydrophobicparticles and the surrounding aqueous macromonomer solution

should lead to more network defects in the vicinity of the parti-cles (Fig. 2).

We first evaluated whether the incorporated hydrophobic NPswould indeed augment network defects and improve permeability.We selected PLGA as a suitable hydrophobic material for the par-ticle because of its biocompatibility (17). When hydrophobic NPs(870� 34 nm diameter) were incorporated into PEG matrices(0.01% wt∕vol) without cells, we observed reduced mechanicalstrength (Fig. 3A), suggesting that the level of network defectshad increased; however, no change was observed when hydrophi-lic NPs (carboxylated polystyrene, 785� 6 nm diameter) wereadded (Fig. 3A). Because albumin (14 nm × 4 nm × 4 nm dimen-sions) is known to be an effective marker to probe effectivenetwork sizes (15), we also compared albumin release from sam-ples containing hydrophobic NPs with release results for samplescontaining hydrophilic NPs. We observed that only hydrophobicNPs improve the permeability of PEG matrices (Fig. 3B); nosuch effect was observed for hydrophilic NP-containing samples(Fig. 3B).

We then incorporated PLGA NPs into cell-encapsulating PEGmatrices. Fig. 3C depicts Huh 7.5 cells encapsulated in PEGmatrices with and without NPs. Cell viability was increased forall the different thicknesses of PLGA NP-containing samples, asshown in Fig. 3D. The increase of cell viability in proportion tothe matrix thickness implies that the dominant benefit fromaddition of PLGA NPs is the improvement in permeability. Wealso evaluated the effects of different conditions of NPs, includinghydrophobicity/hydrophilicity, size, concentration, and PLGAcomposition (Fig. S1). These studies demonstrated that the pre-sence of hydrophobicity of PLGA NPs is the primary factor con-tributing to the improvement in encapsulated cell viability.Fig. 3E shows a significant increase in the live cell population(green fluorescence) and in the number of cell colonies (whitearrows) for the PLGA NP-containing samples, as stained bythe Live/Dead® viability measurement kit. The formation of cellcolonies is important for hepatic cells to function normally be-cause of their dependence on cellular interactions. Thus, our net-work design provides a more desirable environment forencapsulated cells to proliferate and restore homotypic cellularinteractions.

In the preceding experiments studying various encapsulatingconditions, we mainly used Huh 7.5, a hepatoma cell, which iswidely used in HCV research to overcome the limited ability ofother liver-derived cells to support HCV replication, in vitro (18).One problem is that it is typically difficult to maintain primaryhuman hepatocytes in standard 2D in vitro cultures in which theyrapidly lose features of advanced differentiation (19). There hasbeen an attempt to encapsulate rat primary hepatocytes in PEGmatrices, but the viability was too poor to perform biologicalassays (20). Thus, there needs to be a better cell-encapsulationsystem for the 3D culture of primary hepatocytes in order to fullybenefit from the potential advantages offered by PEG hydrogels.Because of the resistance to cell attachment to the PEG networkand the anchor dependence of liver-derived cells, we conjugatedRGD (Arg-Gly-Asp) peptides, a cell binding domain (21), to thePEG network for human primary hepatocyte encapsulation. Totest the benefit of incorporating PLGA NPs, we measured theviabilities of different sample groups. As expected, there weresignificant improvements in viability for NP-containing samples(Fig. 4A).

Because our approach was designed to improve phenotypestability by increasing matrix permeability, we selected urea as ahepatic function marker because the diameter (less than 0.5 nm)is small enough not to be trapped by the network. We verified thatthere was no significant difference in urea diffusion betweensamples with and without PLGA NPs (Fig. S2). As expected, weobserved significant increases in urea secretion, normalized byeach sample’s viability, in the samples with PLGA NPs (Fig. 4B).

0

20

40

60

80

100

120

1 3 5 7 10 14

20% 3.4k PEG (reference)30% 3.4k PEG40% 3.4k PEG

Rel

ativ

e vi

abili

ty (

%)

Time (d)

0

20

40

60

80

100

120

140

1 3 5 7 10 14

20% 3.4k PEG (reference)10% 3.4k PEG20 % 8k PEG10% 8k PEG

Time (d)

Rel

ativ

e vi

abili

ty (

%)

BA

Fig. 1. (A) As a reference, we used 3.4 k molecular weight of PEG andprepared samples from prepolymer with 20% ðwt∕wtÞ solid content, referredto as 20% 3.4 k PEG. When samples were prepared by increasing molecularweight or reducing solid content from 20% 3.4 k PEG, there were no signifi-cant changes in the viability of encapsulated cells compared with 20% 3.4 kPEG (p value ¼ 0.32). (B) However, there were significant cell viability dropsfrom 20% 3.4 k PEG to 30% 3.4 k PEG (p value < 0.01), from 30% 3.4 k PEG to40% 3.4 k PEG (p value < 0.01).

2 of 6 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1005211107 Lee et al.

Dow

nloa

ded

by g

uest

on

July

20,

202

1

We also measured the induction activity of cytochrome P-450monooxygenase 3A4 (CYP3A4). CYP3A4 is involved in the oxi-dative metabolism of various xenobiotics, including more than50% of clinically used drugs (22). The expression level of thisenzyme is related to the accumulation of toxic metabolites, drugside effects, and the therapeutic efficacy of a coadministered drug(23). The induction activity of CYP3A4 was evaluated by therelative expression level of the enzyme between samples culturedwith and without inducers, which avoids permeability effectsbetween samples. We tested induction of CYP3A4 activity torifampicin, a bactericidal antibiotic drug of the rifamycin group,which is known to be one of the most reliable CYP3A4 inducers(24). We used human adult primary hepatocytes from two donorsthat had lost the CYP3A4 induction activity when they were cul-tured on collagen-coated 2D tissue culture plates (Fig. 4C). Eventhough the cells in 3D matrices without NPs showed no responseto the inducer (Fig. 4C), we observed restored induction activityof CYP3A4 in NP-containing samples (Fig. 4C). Taken together,these results show that incorporation of hydrophobic NPs is an

effective way to improve phenotype stability of encapsulatedcells.

New Network Design Maintains Patternability of Cell-EncapsulatingPEG Hydrogels. One of the critical features of engineered tissuescaffolds is the replication of in vivo geometry and dimensionalsize scale in order to provide an environment that could poten-tially guide cell behavior with respect to morphology, cytoskeletalstructure, and functionality (25). In order to obtain more reliablemicroscale architectures within a PEG matrix, there should beappropriate mechanical strength for long-term structural stability,and the viscosity of the fluid prepolymer needs to be low enoughfor the prepolymer to flow into microscale mold structures toallow fabrication through soft lithography or microfluidic proces-sing. However, because the diffusion condition in PEG-basedscaffolds is the primary factor for phenotype stability of encapsu-lated cells, it is a challenge to design a network to satisfy thesecontradictory requirements. Because our network design forimproving the hydrogel permeability was intended to partiallyinduce loose cross-linking only at the particle-PEG interface,

< The actual network structure> < The modified network structure> < The uniformly cross-linked ideal network>

Fig. 2. The distance between consecutive cross-links (red arrow) controls the permeability of polymeric networks only in the case when the network isuniformly cross-linked. However, under the curing condition that was optimized cell-encapsulating PEG networks, there are numerous network defects (bluearrows), which appear to be up to orders ofmagnitude larger than the possible range of distances between consecutive cross-links. Therefore, the level of thesenetwork defects is the primary determinant of the hydrogel permeability and its consequent ability to support metabolic activities of encapsulated cells. Thesenetwork defects could be augmented by the incorporation of hydrophobic NPs (green circles), which are believed to induce even looser cross-linking within thehydrophilic PEG network.

A B

D E

C

Fig. 3. (A) Compressionmodulus of PEGmatrices without NPs, with hydrophilic NPs (polystyrene beads with carboxylated surfaces), or with hydrophobic PLGANPs were measured, and only hydrophobic NP-containing samples had reduced compression modulus (p value < 0.01). (B) When albumin was encapsulated inPEG matrices without cells, samples containing hydrophobic PLGA NPs showed significantly improved albumin release, (p value < 0.01), whereas hydrophilicNPs had no effect on release profiles (p value ¼ 0.57), compared with samples without NPs. (C) PLGA NPs (red arrows) were incorporated within matricesencapsulating Huh 7.5 cells (green arrows). Scale bar: 10 μm. (D) For different thicknesses of PLGA NP-containing samples, increased cell viabilities wereobserved compared to samples without NPs. (E) The improved viabilities in NP-containing samples are shown in the figures stained by Live/Dead® assay (green:live cells; red: dead cells). White arrows indicate cell colonies resulting from proliferation of individual cells. Scale bar: 100 μm.

Lee et al. PNAS Early Edition ∣ 3 of 6

ENGINEE

RING

APP

LIED

BIOLO

GICAL

SCIENCE

S

Dow

nloa

ded

by g

uest

on

July

20,

202

1

we anticipated that this approach would also minimize loss ofphysical properties for patternability.

In order to evaluate the physical characteristics necessary topatternability, we measured the viscosity of the prepolymer andcompression modulus of the cured matrices. As the macromono-mer molecular weight increased from 3.4 k to 8 k, the viscosityincreased (306� 3%) and the compression modulus of the curedgel decreased (77� 4%), as shown in Fig. 5A. Because 8 k PEGmatrices had similar cell viability as for 3.4 k PEG matrices, asshown in Fig. 1A the ideal network mesh size was not an appro-priate design parameter. PLGA NP-containing samples showedno significant change in the viscosity of the prepolymer solution(p value > 0.05) and only modest loss in the gel compressionmodulus (31� 3%) (Fig. 5A). The ability to support increased cellviability while maintaining patternability opens the door to devel-oping hydrogel platforms with more sophisticated architecturesthat promote phenotype stability for different primary cell types.

Ultimately, one condition for phenotype stability is simply highcell viability. With this in mind, one notable feature of the Huh7.5 cells encapsulated in the reference PEG matrices—initiallyused to optimize cell encapsulation conditions—is that the cellviability was always greater at the edges of the hydrogel matrixthan in the middle portion, as reported for rat hepatocytes (13)and human mesenchymal stem cells (26). Because cells requirethe diffusion of nutrients, oxygen, and waste by-products throughthe PEG matrix to support their metabolic activities, we hypothe-sized that lower cell viability in the middle portion was caused byinsufficient permeability of the PEG hydrogel. This correlationbetween cell viability and matrix permeability was confirmed bythe observation that increasing the thickness of the PEG hydrogelled to decreased viabilities in both Huh 7.5 cells and human fetal

B

A

C

Fig. 4. (A) The addition of PLGA NPs led to further improvement of theviability of human fetal primary hepatocytes (p value < 0.01) and humanadult primary hepatocytes (p value < 0.01). (*The viability represents “livecell number relative to the number of initially encapsulated cells.”) (B) Ureasecretion from human adult primary hepatocytes was used as a hepaticfunction marker and normalized by cell viability. Increases in normalizedurea secretion were observed for PLGA NP-incorporated samples (p value <0.01). (C) The CYP3A4 expression levels of human adult primary hepatocyteswere measured and normalized by each sample’s viability. Induction ofCYP3A4 activity of human adult primary hepatocytes was lost for 2D(cultured on collagen-coated plates with 100% confluency, p value ¼ 0.092)and 3D samples without NPs (encapsulated in PEG hydrogels withoutPLGA NPs, p value ¼ 0.73). However, there was significant restoration ofthe induction of CYP3A4 activity in 3D samples with NPs (encapsulated inPLGA nanoparticle-incorporated hydrogels, p value < 0.01).

A B

DC

Fig. 5. (A) The addition of PLGA NPs had no significant effect on viscosity of aqueous 20% 3.4 k PEG solution (p value ¼ 0.51), whereas 20% 8 k PEG had306� 3% higher viscosity than 20% 3.4 k PEG. Samples containing NPs had 31� 3% reduced compression modulus compared to samples without NPs,whereas samples prepared with 8 k PEG had 77� 4% reduced compression modulus compared to samples prepared with 3.4 k PEG. (B) Schematic illus-tration of cell patterning process. Microstructures were fabricated in a silicon wafer by standard microfabrication techniques and then transferred to aPDMS replica. The PEG prepolymer with cells was cured between a glass slide and the PDMS replica. Another type of cell was loaded on the surface, andthey were trapped into the patterned depressions by mild agitation or flow. (C) Fluorescence images of encapsulated cells (red) in PEG matrices andpatterned cells (green) on the surface. Hepatocytes and fibroblasts were encapsulated following 2-d coculturing on cell culture dishes, and fibroblastswere loaded in the patterned depressions on the surface. All samples had the same number of encapsulated hepatocytes and fibroblasts. Type 2 andType 3 had twice the number of fibroblasts on the surface than Type 1. Scale bar: 1.000 μm (D) As more fibroblasts on the surfaces were loaded fromsamples without fibroblasts to sample Type 2, urea secretions increased (p value < 0.01, regression analysis). Type 3 had greater urea secretion than Type 2(p value < 0.01), even though both had the same number of fibroblasts on the surface. All statistical analyses in D were done after a stabilization period(from 5 d after encapsulation).

4 of 6 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1005211107 Lee et al.

Dow

nloa

ded

by g

uest

on

July

20,

202

1

primary hepatocytes (Fig. S3). Thus, patternability can be poten-tially exploited to further increase cell viability.

Herein, however, we primarily focused on controlling thespatial distribution of incorporated cells within cell-encapsulatingPEG in order to demonstrate the potential of our system forreplicating tissue microarchitectures. In many tissues, differenttypes of cells distribute with specific configurations, and theirinteractions are of fundamental importance in physiology, patho-physiology, oncology, developmental biology, and wound healing(27). In the actual liver, for example, hepatocytes are aggregatedin sheets and separated by blood channels, called sinusoids. Theproper distribution of nonparenchymal cells along the sinusoidis a prerequisite to restore various functions of the liver (28).Several approaches have been successfully proposed to restoreheterotypic interactions (29).

We sought a method to guide cells’ spatial displacement with asimplified patterning process, which is given in Fig. 5B, as proof ofthe principle. The patterning process did not hinder cell viability(Fig. S4). Indeed, due to the minimal loss of mechanical strength,the patterned structure was maintained for more than 1 mo. Themicrostructure was designed to control the spatial distributionof encapsulated cells and to trap other types of cells within thepatterned, depressed regions on the surface of cell-encapsulatingPEG matrices (Fig. 5C). Because the incorporation of hydropho-bic NPs was verified to improve homotypic interactions, we rea-soned that we might be able to achieve an even more desirable3D environment for the encapsulated cells by providing a secondcell type on the surface, thus providing both homotypic and het-erotypic cellular interactions, as present in many tissues. Becausecellular interaction between hepatocytes and fibroblasts has beenwell-investigated and is known to enhance hepatic functions (27),we selected fibroblasts for incorporation into our system. It hasbeen reported that cellular interactions between hepatocytesand fibroblasts are initiated only by direct contact-mediatedsignals, but they could be maintained within a short range(∼325 μm) by soluble signals (30). Thus, even after the fibroblastshad been exchanged with naive fibroblasts that had not beenin direct contact with hepatocytes, the hepatocytes could concei-vably maintain the initiated interaction with the naive fibroblastsby soluble signals (30). In order for encapsulated hepatocytesto interact with fibroblasts on the matrix surface, we first cocul-tured human adult primary hepatocytes with NIH 3T3 fibroblastsfor 2 d and encapsulated both cell types into PEG matrices. ThenNIH 3T3 fibroblasts were loaded on the matrix surface. Becausewe had verified that increased interaction with fibroblasts led toincreased secretion of urea and albumin from hepatocytes(Fig. S5), the cellular interaction between encapsulated hepato-cytes and fibroblasts on the matrix surface could be evaluatedby the secreted amounts of these hepatic function markers. In oursystem, the cellular interaction between two types of cells couldbe reliably regulated by controlling cell numbers on the matrixsurface at specific configurations. In Fig. 5C, all samples had thesame number of encapsulated hepatocytes and fibroblasts, andType 2 and Type 3 configurations had twice as many fibroblastson the surface compared to the Type 1 configuration.

As reported in other work (27), it took several days for the twotypes of cells to establish stable heterotypic interactions, so wewaited 5 d after encapsulation before we started examining thesecretion of hepatic markers. As expected, increased numbersof fibroblasts on the matrix surface led to enhanced secretionof urea, as shown in comparison between samples without fibro-blasts on the matrix surface vs. sample Type 1 (Fig. 5D, p value <0.05), and for Type 1 vs. Type 2 (Fig. 5D, p value < 0.01). We alsoobserved that different configurations of patterned cells on thematrix surface resulted in different levels of urea secretion, asshown in comparison of Type 2 and Type 3 (Fig. 5D, p value <0.01). Because the interaction between encapsulated hepatocytesand fibroblasts was mediated by soluble signals, we suggest that

the increased urea secretion for Type 3 compared to Type 2 couldbe due to a more favorable surface configuration in Type 3, inwhich fibroblasts were well-distributed over the surface so thatsoluble signals could be delivered to more encapsulated cells.Even though this experiment was designed only to provide aqualitative evaluation of heterotypic interactions, this approachprovides a platform to understand how distributions of differenttypes of cells and their consequent interactions affect variousbiological responses.

In summary, we first suggested that network defects serve asan important parameter in modifying the physical configurationof the cell-encapsulating PEG network. Based on this hypothesis,we incorporated hydrophobic NPs to augment network defectsby purposefully inducing loose cross-linking at the particle-PEG interface. This approach was verified to be an effective wayto satisfy the seemingly contradictory requirements for an opti-mal network; it improved the permeability to support metabolicactivities of cell-encapsulating PEG matrices while maintainingpatternability by minimizing the loss of mechanical strengthand viscosity. This structural advantage allowed us to constructmicron-scale cell-trapping architectures for the encapsulatedcells to potentially restore heterotypic cellular interactions moreakin to native tissues. We expect that this strategy to designnetwork structures of cell-encapsulating hydrogels can be appliedto restore more natural 3D environments of many tissues withcomplicated architectures.

Materials and MethodsAdditional Materials and Methods are provided in SI Materials and Methods.

Preparation of Hydrophobic Particles of PLGA. A solution of 1% ðwt∕volÞ 50∕50PLGA (composed of 50∕50 molar ratio of glycolide units and lactide units,85 k molecular weight, LACTEL) in dichloromethane (DCM, Sigma-Aldrich)was poured into 100 times its volume of water. The solution was emulsifiedfor 2 min at ∼30;000 rpm using a PRO200 Laboratory Homogenizer (ProScientific). The emulsion was stirred at 1,000 rpm overnight at room tempera-ture and atmospheric pressure to allow DCM to evaporate. The particlesuspension was separated by centrifugation and placed in a vacuum desicca-tor overnight. The average diameter of the NPs, measured by dynamic lightscattering, was 870� 34 nm.

Cell Encapsulation in PEG Matrices. PEG diacrylate (PEG-DA) was dissolved inPBS with 0.05% ðwt∕volÞ of the photoinitiator, Irgacure 2959 (Ciba). Hydro-phobic PLGA NPs were added to the solution for some samples with thedesignated concentrations. For hydrophilic particles, we used 785� 6 nmdiameter polystyrene beads (Polybead® Carboxylate 0.75 Micron Micro-spheres, Polysciences) that had been surface-modified with carboxyl groups.The cell suspension was then added to a given PEG-DA prepolymer solutionat 10 × 106 cells∕mL for Huh 7.5 cells and 30 × 106 cells∕mL for human pri-mary hepatocytes. For primary hepatocyte encapsulation, RGD-conjugatedPEG was mixed with the prepolymer at 20 μmol∕mL. The final prepolymersolution was loaded between glass slides (VWR) separated by a Teflon spacerand exposed to 320–390 nm UV light with 10 mW∕cm2 intensity (Cure spot50, DYMAX) for 50 s. The cured hydrogels were subsequently washedwith PBS and then cultured in ultralow attachmentmultiwell plates (Corning)with cell culture medium; the medium was changed every 2 d. For primaryhepatocyte encapsulation, RGD-conjugated PEG was mixed with the prepo-lymer at 20 μmol∕mL. Note that the curing conditions for cell-encapsulatingPEG matrices involved reducing the photoinitiator concentration, UV expo-sure time, and UV intensity for better cell viability as compared to samplesprepared under much more complete curing conditions used in our previousstudy (31).

Cell Patterning Within Cell-Encapsulating PEG Matrices. Microstructures insilicon wafers were prepared by standard microfabrication processing tech-niques. We created structures 200 μm deep by using the Surface TechnologySystems Deep RIE Etcher. The microstructures could be transferred to poly(dimethylsiloxane) (PDMS) replicas by curing PDMS prepolymer (SYLGARD®184 Silicone Elastomer kit, Dow Corning Corporation) for 30 min at 70 °C overthe silicon wafer. Cell-containing PEG prepolymer was cured between thePDMS replicas and a glass slide. The second type of cells was loaded withinthe microstructure on the PEG matrix surface. By generating slight agitation

Lee et al. PNAS Early Edition ∣ 5 of 6

ENGINEE

RING

APP

LIED

BIOLO

GICAL

SCIENCE

S

Dow

nloa

ded

by g

uest

on

July

20,

202

1

or flow, loaded cells were stably entrapped within the patterned depressionswhere less shear stress was applied. After allowing 1 h for cells to attachon the surface of RGD-conjugated PEG matrices, unattached cells wereremoved by rinsing with cell culture media, and the matrices were incubatedunder normal cell culturing conditions. To obtain images of patternedcells, encapsulated cells were stained red by DiI (Vybrant™ Cell-LabelingSolutions, Molecular Probes®), and patterned cells were stained green byDiO (Vybrant™ Cell-Labeling Solutions, Molecular Probes®).

Statistical Analysis. All measurements in this manuscript were repeated atleast two times to confirm the reproducibility. Experiments done with Huh7.5 cells had n ≥ 6, and experiments done with human primary hepatocyteshad n ≥ 3. Statistical analysis was performed by One-way ANOVA withBonferroni–Holm post hoc test through overall time points. Regression

analysis was performed through overall time points with average values ofeach sample. All values are reported as the mean and standard deviation ofthe mean.

ACKNOWLEDGMENTS. The authors gratefully acknowledge the graduatefellowship of the Mogam Scientific Scholarship (to W.L.), the AmericanLiver Foundation Postdoctoral Fellowship Award and a Stanford Dean’sPostdoctoral Fellowship (to N.-J.C.), a Beckman Interdisciplinary TranslationalResearch Program Award, a Burroughs Welcome Fund Clinical ScientistAward in Translational Research (to J.S.G.), the Center for TranslationalResearch in Chronic Viral Infections, the Center on Polymer Interfaces andMacromolecular Assemblies, and a National Science Foundation MaterialsResearch Science and Engineering Center.

1. Hubbell JA (1995) Biomaterials in tissue engineering. Nat Biotechnol 13:565–576.2. Cushing MC, Anseth KS (2007) Hydrogel cell cultures. Science 316:1133–1134.3. Langer R, Vacanti JP (1993) Tissue engineering. Science 260:920–926.4. Jhon MS, Andrade JD (1973) Water and hydrogels. J Biomed Mater Res 7:509–522.5. Sawhney AS, Pathak CP, Hubbell JA (1993) Interfacial photopolymerization of

poly(ethylene glycol)-based hydrogels upon alginate-poly(l-lysine) microcapsules forenhanced biocompatibility. Biomaterials 14:1008–1016.

6. Drury JL, Mooney DJ (2003) Hydrogels for tissue engineering: scaffold design variablesand applications. Biomaterials 24:4337–4351.

7. Khademhosseini A, Langer R, Borenstein J, Vacanti JP (2006) Microscale technologiesfor tissue engineering and biology. Proc Natl Acad Sci USA 103:2480–2487.

8. Sachlos E, Czernuszka JT (2003) Making tissue engineering scaffolds work. Review:The application of solid freeform fabrication technology to the production of tissueengineering scaffolds. Eur Cells Mater 5:29–39.

9. Lee KY, Mooney DJ (2001) Hydrogels for tissue engineering. Chem Rev 101:1869–1880.10. de Gennes PG (1979) Scaling Concepts in Polymer Physics (Cornell Univ Press,

Ithica, NY).11. Cruise GM, et al. (1999) In vitro and in vivo performance of porcine islets encapsulated

in interfacially photopolymerized poly (ethylene glycol) diacrylate membranes. CellTransplant 8:293–306.

12. Cho NJ, et al. (2009) Viral infection of human progenitor and liver-derived cells encap-sulated in three-dimensional PEG-based hydrogel. Biomed Mater 4:11001–11007.

13. Tsang VL, et al. (2007) Fabrication of 3D hepatic tissues by additive photopatterningof cellular hydrogels. FASEB J 21:790–801.

14. Cruise GM, Scharp DS, Hubbell JA (1998) Characterization of permeability andnetwork structure of interfacially photopolymerized poly(ethylene glycol) diacrylatehydrogels. Biomaterials 19:1287–1294.

15. Gong JP, et al. (2001) Synthesis of hydrogels with extremely low surface friction. J AmChem Soc 123:5582–5583.

16. Gong JP, Kii A, Xu J, Hattori Y, Osada Y (2001) A possible mechanism for the substrateeffect on hydrogel formation. J Phys Chem B 105:4572–4576.

17. Lee W, Wiseman ME, Cho NJ, Glenn JS, Frank CW (2009) The reliable targeting ofspecific drug release profiles by integrating arrays of different albumin-encapsulatedmicrosphere types. Biomaterials 30:6648–6654.

18. Blight KJ, McKeating JA, Rice CM (2002) Highly permissive cell lines for subgenomicand genomic hepatitis C virus RNA replication. J Virol 76:13001–13014.

19. Maher JJ (1988) Primary hepatocyte culture: Is it home away from home? Hepatology8:1162–1166.

20. Underhill GH, Chen AA, Albrecht DR, Bhatia SN (2007) Assessment of hepatocellularfunction within PEG hydrogels. Biomaterials 28:256–270.

21. Ruoslahti E, Pierschbacher MD (1986) Arg-Gly-Asp: A versatile cell recognition signal.Cell 44:517–518.

22. Lehmann JM, et al. (1998) The human orphan nuclear receptor PXR is activatedby compounds that regulate CYP3A4 gene expression and cause drug interactions.J Clin Invest 102:1016–1023.

23. Moore LB, et al. (2000) Orphan nuclear receptors constitutive androstane receptorand pregnane X receptor share xenobiotic and steroid ligands. J Biol Chem275:15122–15127.

24. Bachmann KA, Jauregui L (1993) Use of single sample clearance estimates of cyto-chrome P450 substrates to characterize human hepatic CYP status in vivo. Xenobiotica23:307–315.

25. Desai TA (2000) Micro-and nanoscale structures for tissue engineering constructs.MedEng Phys 22:595–606.

26. Nuttelman CR, Tripodi MC, Anseth KS (2004) In vitro osteogenic differentiation ofhuman mesenchymal stem cells photoencapsulated in PEG hydrogels. J BiomedMater Res 68:773–782.

27. Bhatia SN, Balis UJ, Yarmush ML, Toner M (1999) Effect of cell-cell interactions in pre-servation of cellular phenotype: Cocultivation of hepatocytes and nonparenchymalcells. FASEB J 13:1883–1900.

28. Jungermann K, Keitzmann T (1996) Zonation of parenchymal and nonparenchymalmetabolism in liver. Annu Rev Nutr 16:179–203.

29. Andersson H, Berg A (2004) Microfabrication and microfluidics for tissue engineering:state of the art and future opportunities. Lab Chip 4:98–103.

30. Hui EE, Bhatia SN (2007) Micromechanical control of cell–cell interactions. Proc NatlAcad Sci USA 104:5722–5726.

31. Myung D, et al. (2008) Glucose-permeable interpenetrating polymer networkhydrogels for corneal implant applications: A pilot study. Curr Eye Res 33:29–43.

6 of 6 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1005211107 Lee et al.

Dow

nloa

ded

by g

uest

on

July

20,

202

1